Molecular Ecology Resources (2010) 10, 677683 doi: 10.1111/j.1755-0998.2009.02824.

TECHNICAL ADVANCES

A rapid column-based ancient DNA extraction method for

increased sample throughput
NADIN ROHLAND,* HEIKE SIEDEL* and M I C H A E L H O F R E I T E R *
*Max Planck Institute for Evolutionary Anthropology, Deutscher Platz 6, D-04103 Leipzig, Germany, Department of Genetics,
Harvard Medical School, Boston, Massachusetts 02115, USA, Department of Biology, University of York, YO10 5YW, York, UK

Abstract
Genetic analyses using museum specimens and ancient DNA from fossil samples are becom-
ing increasingly important in phylogenetic and especially population genetic studies. Recent
progress in ancient DNA sequencing technologies has substantially increased DNA sequence
yields and, in combination with barcoding methods, has enabled large-scale studies using
any type of DNA. Moreover, more and more studies now use nuclear DNA sequences in
addition to mitochondrial ones. Unfortunately, nuclear DNA is, due to its much lower copy
number in living cells compared to mitochondrial DNA, much more difficult to obtain from
low-quality samples. Therefore, a DNA extraction method that optimizes DNA yields from
low-quality samples and at the same time allows processing many samples within a short
time frame is immediately required. In fact, the major bottleneck in the analysis process
using samples containing low amounts of degraded DNA now lies in the extraction of sam-
ples, as column-based methods using commercial kits are fast but have proven to give very
low yields, while more efficient methods are generally very time-consuming. Here, we pres-
ent a method that combines the high DNA yield of batch-based silica extraction with the
time-efficiency of column-based methods. Our results on Pleistocene cave bear samples show
that DNA yields are quantitatively comparable, and in fact even slightly better than with
silica batch extraction, while at the same time the number of samples that can conveniently
be processed in parallel increases and both bench time and costs decrease using this method.
Thus, this method is suited for harvesting the power of high-throughput sequencing using
the DNA preserved in the millions of paleontological and museums specimens.
Keywords: ancient DNA, columns, extraction, museum specimens, silica

Received 2 September 2009; revision received 18 November 2009; accepted 29 November 2009

similar trend is taking place in the analysis of museum

Introduction
The field of ancient DNA has grown substantially during
the last decade, moving away from the analysis of indi-
vidual samples in phylogenetic analyses to population
genetic analyses (Leonard et al. 2000) that require dozens
(Barnes et al. 2002; Hofreiter et al. 2004; Weinstock et al.
2005) and sometimes hundreds of samples (Shapiro et al.
2004). This and the emerging fields like palaeogenomics
(Green et al. 2006; Poinar et al. 2006; Miller et al. 2008)
and phenotypic analyses using nuclear single-nucleotide
polymorphisms (Ludwig et al. 2009) require testing many
samples (Pennisi 2009) and optimizing DNA yields. A
Correspondence: Nadin Rohland, Fax: (617) 432-7663;
E-mail: nrohland@genetics.med.harvard.edu
specimens, which are becoming increasingly important
for genetic analyses (Wandeler et al. 2007) with
population genetic studies now often being extensively
(Godoy et al. 2004) or even exclusively (Miller et al. 2006;
Krystufek et al. 2007) based on museum specimens.
Finally, in forensic science, DNA analyses are, among
others, used for the identification of criminals and
victims of crime or war. Often such analyses also have to
be performed on degraded bone or teeth samples
(Iwamura et al. 2004), and when dealing with war or
natural disaster victims, analyses have to be performed
on large numbers of degraded samples (Holland et al.
2003; Deng et al. 2005; Lehrman 2006).
Like many fields in molecular biology, research on
ancient and museum specimen DNA is becoming

 2009 Blackwell Publishing Ltd

678 T E C H N I C A L A D V A N C E S
completely transformed by the introduction of high-
throughput DNA-sequencing methods (Margulies et al.
2005; Schuster 2008), often referred to as next-generation
sequencing (NGS). The large number of reads obtained
and the fact that each read represents a clonal sequence
make them ideally suited for ancient DNA research
(Gilbert et al. 2007; Miller et al. 2008), whereas the short
sequencing length of NGS is no disadvantage given the
fragmented nature of ancient DNA templates (Paabo
et al. 2004; Poinar et al. 2006). Moreover, barcoding meth-
ods (Meyer et al. 2007; Erlich et al. 2009) allow sequencing
of hundreds of specimens in a single-sequencing run
raising the possibility of large-scale population studies
using ancient or museum specimens.
However, to harvest the power of NGS technology for
the analysis of ancient and historical specimens, a DNA
extraction methodology that is efficient with regard to
both DNA yields and the number of samples that can be
processed in parallel is immediately required. Unfortu-
nately, commercial column-based methods give insuffi-
cient yields of DNA (Rohland & Hofreiter 2007b) and,
moreover, are usually not amenable to use with large
sample volumes. Silica- (Rohland & Hofreiter 2007b),
precipitation- (Vigilant et al. 2001) and microfilter-based
methods (Leonard et al. 2000; Schwarz et al. 2009) are
time-consuming and labour-intensive. In fact, the extrac-
tion of DNA from low DNA quality and quantity sam-
ples is now the rate-limiting step in the data production
process using such samples. Therefore, we set out to opti-
mize and combine the high DNA yields of a previously
published silica extraction method with the more conve-
nient and faster handling of column-based extractions.
Materials and methods
We used two to three Pleistocene cave bear samples
(samples e, g and i; for sample information, see Table S1)
from three different caves in Austria for initial compari-
sons of a variety of parameters during DNA extraction.
The DNA extraction protocol, referred to as batch
method, we started with (Rohland & Hofreiter 2007a)
uses a simple extraction buffer (consisting of 0.45 M
EDTA, pH 8.0 and 0.25 mg mL proteinase K) to deminer-
alize and digest bone or tooth powder. After overnight
incubation at RT, the supernatant (1 volume) is added to
four times the volume of binding buffer [contrary to the
protocol in Rohland & Hofreiter (2007a,b), the binding
buffer we used consists solely of 5 M guanidinium thiocy-
anate (GuSCN)] and 100 lL of silica suspension, and the
pH is adjusted to 4 with hydrochloric acid. This mix-
ture is incubated for 3 h under agitation. During this
incubation, DNA molecules bind to the silica surface.
Subsequent resuspensions of the silica pellet with wash-
ing buffer (50% ethanol, 125 mM NaCl, 1 TE) are
required to remove salts and PCR-inhibiting agents,
which get co-extracted from the specimen. Finally, the sil-
ica pellet is dried and DNA is eluted with 1 TE.
We investigated modifications of the method at all
three stages of the protocol, the sample digestion step,
the DNA binding step to silica and, finally, the washing
step.
First, we tested whether it is possible to reduce the
volume of extraction buffer used, as extremely large vol-
umes are difficult to handle with flow-through columns.
Therefore, we tested the DNA yield per gram of sample
powder using different ratios of sample powder to buffer
volume, varying from 50 to 400 mg mL in a total volume
of 5 mL.
Second, we tested whether it is possible to change the
ratio of binding buffer to extraction buffer. We initially
tested ratios of binding to extraction buffer ranging from
the initial ratio of 4:1 down to 1:2. Due to the results of
this test, we did a further set of tests using ratios between
1:1 and 1:10. In all experiments, we used 100 lL of silica
suspension. To make the protocol more robust against
handling variation, we also eliminated the pH adjust-
ment step after combining binding buffer, silica and
extraction buffer by buffering the binding buffer using
sodium acetate (pH 5.2) with a final concentration in the
binding buffer of 300 mM. Thus, the binding buffer now
solely consists of 5 M GuSCN and 300 mM sodium ace-
tate, pH 5.2. The incubation of extraction buffer together
with binding buffer and silica is still performed in sus-
pension for 3 h as we previously found that this length of
time is necessary for optimal results and shorter incuba-
tion times result in a reduction of DNA yields (Rohland
& Hofreiter 2007b).
Third, to simplify all subsequent washing and elution
steps, we then immobilize the silica particles on a filter
(e.g. glass microfibre binder-free Grade GF B: 1.0 lm
Fig. 1 Set up of the columns: Mobicols can be ordered with
the 10 lm filter already applied to the column (MobiTec,
Germany, product # M1002S). A filter with smaller pore size
(e.g. Whatman GF/B 1.0 lm) needs to be placed on top of the
large filter to prevent the small silica particles from being
washed through during vacuum or centrifugation steps.
 2009 Blackwell Publishing Ltd

from Whatman) that can be fitted into commercially
available columns (Mobicols with 10 lm filter, M1002S;
MoBiTec GmbH) by using a 7-mm hole punch (see Fig. 1
for column assembly). The addition of the 1-lm filter is
necessary because the silica particles used to bind DNA
are up to 10 lm in diameter and the majority would
therefore not be retained by the 10-lm filter already
applied to the columns. The silica suspension from the
binding step, including extraction and binding buffer,
is centrifuged for 2 min at 5000 g, the supernatant is
discarded and the silica pellet is resuspended in 400 lL
binding buffer. The suspension is loaded on top of the
filter columns and the columns are applied to a vacuum
manifold. Washing of the silica, to remove inhibitors
originating from the specimens as well as traces of
GuSCN, is carried out at least twice using 450 lL
of washing buffer by adding the washing buffer on top of
the silica and applying vacuum. We recommend a short
centrifugation step of the columns before the first and
after the last washing step to remove any remaining salt
and ethanol, which could lead to the inhibition of down-
stream applications. After this centrifugation step, DNA
is eluted by pipetting 1 TE on top of the silica layer,
short incubation and subsequent centrifugation into a
new labelled tube.
A detailed description of the new protocol can be
found in the Supporting Information.
The performance of the different extraction variants
investigated was tested using quantitative real-time PCR
(qPCR) on a 110-bp fragment of the mitochondrial DNA
with a cave bear-specific fluorescently labelled TaqMan
probe exactly as in Rohland & Hofreiter (2007b). Usually
the copy number was measured from two different
concentrations of the extract (undiluted and 1:10 dilu-
tion) in two independent amplifications per concentra-
tion; results with standard deviations are given in
Tables S2S4. The effect of different parameters on
extraction performance (copy number per gram) com-
pared with the best performing set up was calculated for
each experiment and a paired Students t-test was
applied to investigate whether the observed differences
are significant.
After we had established the protocol for column-
based extraction, we tested its performance on two dif-
ferent sets of samples. First, we tested the method in
direct comparison by qPCR to a silica batch method,
which we had found previously to give optimal results
with regard to ancient DNA yields (Rohland & Hofreiter
2007b). This comparison was performed on nine Pleisto-
cene cave bear bone and teeth samples in duplication.
Second, we extracted DNA from 13 chimpanzee teeth
museum samples with a specimen age ranging from 4 to
22 years. These samples derive from chimpanzees that
died in the wild and have been buried for c. 1 year at the
 2009 Blackwell Publishing Ltd
T E C H N I C A L A D V A N C E S 679
field site for tissue removal. We attempted amplification
of nuclear markers, the amelogenin locus as well as
19 microsatellite markers described for chimpanzees,
ranging in length from 104 to 256 bp (Arandjelovic et al.
2009). From all teeth, we cut off a 130- to 230-mg piece of
the tooth root and ground it into a fine powder using a
freezer mill. For these samples, DTT and Triton X-100
were added to the extraction buffer in final concentra-
tions of 50 mM and 1% respectively. Due to the value of
these samples, a direct comparison using an alternative
extraction method was not possible. However, it should
be noted that three of the samples had been extracted
previously using a different method (Vigilant et al. 2001)
and amplification of microsatellite markers had been
unsuccessful.
Results
We found that it is not possible to reduce the extraction
buffer volume (0.45 M EDTA, pH 8, 0.25 mg mL protein-
ase K), as both the DNA yield per gram of bone powder
and the total DNA yield per extraction decrease if more
than 50 mg of bone powder is used per 1 mL of extrac-
tion buffer (Table 1). When qPCRs on further dilutions
of the same extracts were performed, the amount of DNA
per gram of bone powder did not increase (Table S2b),
as it would be expected if the decreasing yield with
higher amounts of bone powder was due to co-extracted
inhibitors.
However, we found that it is possible to reduce the
ratio of binding buffer (5 M GuSCN) to extraction buffer
from the ratio of 4:1 as usually used. Although the DNA
yields remain constant from a ratio of 4:1 to 1:1 of binding
Table 1 Results of qPCR comparing different amounts of
sample powder in the extraction buffer (kept constant at 5ml).
Absolute copy numbers (from two measurements of two
dilutions, each) per gram of powder are shown together with
relative numbers (in parenthesize) compared to the best
performing method for each sample in this experiment.
Significantly worse performing input sample amounts
compared to the best performing ratio are marked with an
asterisk (*, P<0.05, paired Students t-test)
Sample ID e g i
Approximate
amount of
sample Average of
powder Copy number per gram relative
(mg mL) (relative to best method) performance
50 29 795 (1) 9 541 (1) 86 (0.22) 0.74
100 4 972 (0.17) 3 889 (0.41) 386 (1) 0.53
200 1 464 (0.05) 1 670 (0.18) 30 (0.08) 0.10*
400 0 (0) 0 (0) 0 (0) 0*
680 T E C H N I C A L A D V A N C E S

to extraction buffer, a ratio of 1 volume binding buffer to
2 volumes extraction buffer gave even better DNA yields.
Below this ratio, DNA yields decrease rapidly (Table 2).
As GuSCN represents the most expensive reagent in the
extraction process, this modification has the added
benefit that it reduces the costs per extraction by 50%
compared with the initial batch method.
We tested the column method using 250 mg bone
powder in 5 mL extraction buffer with a ratio of binding
buffer (5 M GuSCN, 300 mM sodium acetate, pH 5.2) to
extraction buffer of 1:2 (2.5 mL binding buffer) and
Table 2 Results of qPCR comparing different ratios of extraction
to binding buffer (amount of extraction buffer kept constant at
5ml). Absolute copy numbers (from two measurements of two
dilutions, each) per gram of powder are shown together with
relative numbers (in parenthesize) compared to the best
performing method for each sample in the respective experiment
(a and b). Significantly worse performing buffer ratios compared
to the best performing ratio are marked with an asterisk
(*, P<0.05, Students t-test). a) initial experiment, varying the
ratio from 4:1 to 1:2. b) second experiment, varying the ratio
from 1:1 to 1:10
binding of DNA to silica for 3 h in suspension, followed
by washing and elution on columns on nine Pleistocene
cave bear bones and teeth in duplication (Table 3). In this
comparison, the new protocol gave results comparable
and in fact even better than our previously published sil-
ica batch method (Rohland & Hofreiter 2007a,b) using
resuspension for all washing and elution steps (Fig. 2
and Table 3). Although we found considerable variation
for some samples, and the overall difference is statisti-
cally not significant, the column-based method was supe-
rior in this test for six of the nine samples.
For the 13 chimpanzee teeth, we were able to amplify
all 19 nuclear microsatellites from six of the samples
(Table S5). A further two samples yielded partial profiles
(15 and 18 loci respectively). For Amelogenin, we found a
total success rate of 61% (8 of 13 samples, all loci ampli-
fied in duplicate). Three of the samples were teeth that
had been previously extracted (Vigilant et al. 2001) and
had failed to amplify. One of them could be amplified for
18 of the 19 loci, one for 15 loci and one failed again using
this method.

Sample ID e g Discussion

Ratio of
The aim of this study was to develop a DNA extraction
binding to Average of method for paleontological and historical (museums
extraction Copy number per gram relative specimens) bone and teeth samples that maximizes DNA
buffer (relative to best method) performance retrieval and reduces the time and workload required.
Both, the comparison between the column-based method
(a) developed and the batch-based silica extraction using
4:1 75 459 (0.27) 14 953 (0.72) 0.5
nine Pleistocene cave bear samples, and the extraction of
3:1 77 986 (0.28) 15 041 (0.73) 0.51
2:1 53 273 (0.19) 17 905 (0.87) 0.53
13 historical chimpanzee teeth, show that the column
1:1 18 657 (0.07) 18 128 (0.88) 0.48 extraction gives consistently high DNA yields. Although
1:2 281 309 (1) 20 686 (1) 1 the column-based and original batch-based methods do
(b) not differ significantly in yields for the cave bear samples,
1:1 20 129 (0.07) 16 887 (0.75) 0.41 this is the first column-based method that, in our experi-
1:1.25 58 540 (0.21) 19 066 (0.85) 0.53 ence, is not worse than batch-based silica extraction, but
1:1.67 268 207 (0.94) 18 968 (0.85) 0.9
rather, if anything, better. It is notable that this method
1:2.5 284 140 (1) 22 437 (1) 1
also works well on younger museum specimens, as
1:5 10 757 (0.04) 1 452 (0.06) 0.05*
1:10 26 (0) 886 (0.04) 0.02* shown by the results on the chimpanzee teeth for nuclear
marker. For such younger samples, we recommend using

Table 3 Results of qPCR comparing the initial extraction technique with the new column-based extraction method on nine cave bear
samples. Absolute copy numbers (from two independent extractions and two measurements of two dilutions, resulting in 8
measurements per sample per method) per gram of powder are shown together with relative numbers (in parenthesize) compared to the
best performing method for each sample in this experiment. The difference between both methods is not significant after a Students
t-test (P<0.05)

Sample ID a b c d e f g h i
Average
Extraction of relative
method Copy number per gram (relative to best method) performance

A initial 63 209 (1) 2 432 (0.52) 1 385 (0.6) 8 331 (0.93) 45 498 (0.1) 165 (1) 9 724 (0.35) 165 (1) 2 027 (0.51) 0.67
B columns 31 091 (0.49) 4 659 (1) 2 293 (1) 8 957 (1) 467 461 (1) 22 (0.13) 28 165 (1) 74 (0.45) 3 966 (1) 0.79

 2009 Blackwell Publishing Ltd


Fig. 2 Relative results of the initial (A)
and the new column-based method (B).
Copy number values are given relative to
the copy number of the best performing
Relative to best method
method for each of the nine cave bear
samples (a-i, see supporting online infor-
mation, Table S1). The horizontal line indi-
cates the average of the respective method
over the nine samples, but is not signifi-
cantly different after a Students t-test
(P<0.05). A, initial extraction method as
previously described (Rohland & Hofrei-
ter, 2007a; Rohland & Hofreiter, 2007b), B,
new column-based method.
DTT and Triton X-100 in the concentrations as described
for the extraction of the chimpanzee teeth. Although we
did not find any beneficial effect on the extraction of
Pleistocene specimens, neither of these reagents had a
negative effect (Rohland & Hofreiter 2007b), and with
younger specimens with more intact tissue structure, the
addition of these two chemicals may have a positive
effect on DNA yields. With regard to such younger speci-
mens, it is especially encouraging that we were able to
extract DNA from samples that had previously failed. In
summary, it is likely that our protocol is suitable for the
extraction of DNA from any degraded hard tissue sam-
ple, including not only fossil and museums specimens
but also historical and forensic samples (Capelli et al.
2003; Holland et al. 2003; Deng et al. 2005; Lehrman 2006;
Coble et al. 2009).
As we showed previously that the batch-based silica
method gives higher DNA yields for ancient samples
than commercially available kits, the modified extraction
protocol presented in this study should outperform these
kits. This difference between different column-based
extractions is most likely explained by the fact that, in
our protocol, binding of DNA is performed for 3 h in sus-
pension whereas with commercial kits DNA can only
bind during the few seconds while it is being washed
through the silica matrix (Rohland & Hofreiter 2007b).
This step provides a major difference to commercial
column-based methods and may at least partially explain
the different yields, as we found previously that an
extended binding time is critical for high DNA yields.
The change in the ratio of binding to extraction buffer
from 4:1 to 1:2 seems to also contribute to the on average
better performance of the column-based extraction com-
pared with the previous protocol (Table 2). An interest-
ing side effect of this change (apart from the slightly
higher DNA yields) lies in the reduced costs, as GuSCN
is the most expensive item in the protocol. Thus,
 2009 Blackwell Publishing Ltd
T E C H N I C A L A D V A N C E S 681
1.0
0.9
0.8 0.79
0.7 0.67
0.6
0.5
0.4
0.3
0.2
0.1
a b c d e f g h i a b c d e f g h i
0.0
A B
Method
compared to the previous batch-based extraction proto-
col, the total costs per extraction of 250 mg bone powder
are reduced approximately by half (already taking
the costs for the columns into account) using the column
protocol.
A further important result of our tests lies in the dis-
covery that adding too much bone powder to a certain
amount of extraction buffer is highly detrimental for the
DNA yield. Interestingly, this was not only the case for
yields per milligram of bone but in fact for total yields
per extraction (Table 1). Quantitative PCR on a dilution
series of the extracts showed that this failure of large
amounts of bone powder to yield DNA is not due to inhi-
bition of the PCR, as up to a dilution of 1:250, extractions
using smaller amounts of bone powder generally yielded
higher copy numbers, both per millilitre of extract and
per gram of bone powder (Table S2b), in accordance with
previous findings on forensic specimens (Loreille et al.
2007). We suggest two mutually nonexclusive explana-
tions for this observation. First, substances inhibiting pro-
teinase K activity may be released. Second, ancient DNA
may mostly be preserved in the inner parts of the bone
powder granules. Thus, at a high bone powder to extrac-
tion buffer ratio, decalcification is insufficient to release
the majority of the DNA in the powder, at least using the
incubation temperatures and times in this protocol. As
EDTA alone has been shown to allow DNA extraction
from ancient bone, albeit with reduced efficiency
(Rohland & Hofreiter 2007b), we believe that the second
mechanism or a combination of both mechanisms is the
more likely explanation. However, further experiments
would be necessary to clarify this issue. In any case, more
is clearly not always better, and researchers should
refrain from using too much bone powder per extraction.
However, using smaller samples has the positive side
effect that damage to specimens is limited; an issue that
is growing in importance due to increasing interest of
682 T E C H N I C A L A D V A N C E S
geneticists in museums collections (Collins et al. 2009;
Nicholls 2009). On the other hand, when larger samples
are available, our method has the advantage that, in con-
trast to commercial kits for which binding to silica is
already performed on the column, it is scalable to almost
any degree. Even if large amounts of extraction and cor-
respondingly binding buffer are used, after the binding
of DNA to the silica, it can be resuspended in a small vol-
ume and applied to the columns.
Although the binding of DNA is still performed in
batch, the use of columns for all steps after the binding
step reduces the workload substantially and reduces the
risk of cross-contamination compared to resuspension
by extensive pipetting. When using standard vacuum
manifolds or table-top centrifuges, in case no vacuum
manifold is accessible, it is possible to process up to 24
samples per extraction. We recommend including at least
two negative controls resulting in 22 specimens that can
be processed in parallel. The original batch method uses
suspension throughout the whole procedure, involving
many time-consuming resuspension steps. Although it is
possible to process up to 24 samples in parallel using the
batch protocol, in our experience just the washing and
elution steps take at least 4 h constant pipetting for an
experienced person, whereas using the new protocol this
part of the protocol requires only c. 1 h for 24 samples, as
no time-consuming resuspension steps of the silica pellet
are necessary anymore. Thus, the bench time required is
reduced substantially. Furthermore, by omitting the
resuspension steps during washing and elution, no
pipetting of DNA-containing solutions is needed, reduc-
ing the possibility of cross-contamination. Finally, a fur-
ther simplification of the protocol lies in the omission of
the pH adjustment after extraction buffer, silica and bind-
ing buffer are combined, and its substitution by buffering
the binding buffer using sodium acetate. Although this
modification results in a pH between 5 and 6, above
the ideal value for DNA binding of pH 4 (Rohland &
Hofreiter 2007b), the results show that this does not harm
DNA yields. We cannot exclude that using a buffer
system at pH 4 would further increase DNA yields. How-
ever, using a buffer system simplifies handling and elimi-
nates the risk of adjusting the pH to below 4, which
results in a rapid loss of DNA (T. Maricic, pers. comm.).
The protocol developed is robust, easy to handle and
combines high DNA yields with reduced bench time per
sample. It also facilitates the parallel processing of larger
sample numbers and is therefore also well suited for har-
vesting the power of NGS, which allow obtaining giga-
bases of DNA sequences in a single-sequencing run, for
samples containing degraded DNA. This is becoming
increasingly important, as various barcoding methods
have been introduced that allow sequencing large num-
bers of specimens in parallel (Meyer et al. 2007; Craig
et al. 2008; Cronn et al. 2008; Illumina 2008; Stiller et al.
2009).
Acknowledgements
We thank Matthias Meyer for encouragement, Holger
Roempler for pointing us to the columns, Tomislav Maricic for
unpublished data on the effect of pH on DNA yields, Kevin
Langergraber (Ngogo, Kibale National Park, Uganda) and
Christophe Boesch (Tai National Park, Ivory Coast) for provid-
ing the chimpanzee samples, Carolyn Rowney for assistance
with the chimpanzee samples, David Blum for help with the
figures, Linda Vigilant and two anonymous reviewers for
comments on the manuscript and the Max Planck Society for
financial support.
References
Arandjelovic M, Guschanski K, Schubert G et al. (2009) Two-step
multiplex polymerase chain reaction improves the speed and
accuracy of genotyping using DNA from noninvasive and
museum samples. Molecular Ecology Resources, 9, 2836.
Barnes I, Matheus P, Shapiro B, Jensen D, Cooper A (2002)
Dynamics of Pleistocene population extinctions in Beringian
brown bears. Science, 295, 22672270.
Capelli C, Tschentscher F, Pascali VL (2003) Ancient protocols
for the crime scene? Similarities and differences between
forensic genetics and ancient DNA analysis. Forensic Science
International, 131, 5964.
Coble MD, Loreille OM, Wadhams MJ et al. (2009) Mystery
solved: the identification of the two missing Romanov chil-
dren using DNA analysis. PLoS ONE, 4, e4838.
Collins MJ, Penkman KE, Rohland N et al. (2009) Is amino acid
racemization a useful tool for screening for ancient DNA in
bone? Proceedings of the Royal Society of London, Series B: Biologi-
cal Sciences, 276, 29712977.
Craig DW, Pearson JV, Szelinger S et al. (2008) Identification
of genetic variants using bar-coded multiplexed sequencing.
Nature Methods, 5, 887893.
Cronn R, Liston A, Parks M et al. (2008) Multiplex sequencing
of plant chloroplast genomes using Solexa sequencing-
by-synthesis technology. Nucleic Acids Research, 36, e122.
Deng YJ, Li YZ, Yu XG et al. (2005) Preliminary DNA identifica-
tion for the tsunami victims in Thailand. Genomics Proteomics
Bioinformatics, 3, 143157.
Erlich Y, Chang K, Gordon A et al. (2009) DNA Sudoku har-
nessing high-throughput sequencing for multiplexed speci-
men analysis. Genome Research, 19, 12431253.
Gilbert MT, Tomsho LP, Rendulic S et al. (2007) Whole-genome
shotgun sequencing of mitochondria from ancient hair shafts.
Science, 317, 19271930.
Godoy JA, Negro JJ, Hiraldo F, Donazar JA (2004) Phylogeogra-
phy, genetic structure and diversity in the endangered
bearded vulture (Gypaetus barbatus, L) as revealed by mito-
chondrial DNA. Molecular Ecology, 13, 371390.
Green RE, Krause J, Ptak SE et al. (2006) Analysis of one million
base pairs of Neanderthal DNA. Nature, 444, 330336.
Hofreiter M, Rabeder G, Jaenicke-Despres V et al. (2004) Evi-
dence for reproductive isolation between cave bear popula-
tions. Current Biology, 14, 4043.
 2009 Blackwell Publishing Ltd

Holland MM, Cave CA, Holland CA, Bille TW (2003) Develop-
ment of a quality, high throughput DNA analysis procedure
for skeletal samples to assist with the identification of victims
from the World Trade Center attacks. Croatian Medical Journal,
44, 264272.
Illumina I (2008) Genome Analyzer Applications: multiplexed
sequencing.
Iwamura ES, Soares-Vieira JA, Munoz DR (2004) Human identi-
fication and analysis of DNA in bones. Revista do Hospital das
Clinicas; Faculdade de Medicina da Universidade de Sao Paulo, 59,
383388.
Krystufek B, Buzan EV, Hutchinson WF, Hanfling B (2007)
Phylogeography of the rare Balkan endemic Martinos vole,
Dinaromys bogdanovi, reveals strong differentiation within the
western Balkan Peninsula. Molecular Ecology, 16, 12211232.
Lehrman S (2006) Missing no longer. Scientific American, 295,
1820.
Leonard JA, Wayne RK, Cooper A (2000) Population genetics of
ice age brown bears. Proceedings of the National Academy of Sci-
ences, USA, 97, 16511654.
Loreille OM, Diegoli TM, Irwin JA, Coble MD, Parsons TJ (2007)
High efficiency DNA extraction from bone by total demineral-
ization. Forensic Science International: Genetics, 1, 191195.
Ludwig A, Pruvost M, Reissmann M et al. (2009) Coat color vari-
ation at the beginning of horse domestication. Science, 324, 485.
Margulies M, Egholm M, Altman WE et al. (2005) Genome
sequencing in microfabricated high-density picolitre reactors.
Nature, 437, 376380.
Meyer M, Stenzel U, Myles S, Prufer K, Hofreiter M (2007)
Targeted high-throughput sequencing of tagged nucleic acid
samples. Nucleic Acids Research, 35, e97.
Miller CR, Waits LP, Joyce P (2006) Phylogeography and mito-
chondrial diversity of extirpated brown bear (Ursus arctos)
populations in the contiguous United States and Mexico.
Molecular Ecology, 15, 44774485.
Miller W, Drautz DI, Ratan A et al. (2008) Sequencing the
nuclear genome of the extinct woolly mammoth. Nature, 456,
387390.
Nicholls H (2009) Time to sequence the red and the dead. Nat-
ure, 458, 812813.
Paabo S, Poinar H, Serre D et al. (2004) Genetic analyses from
ancient DNA. Annual Review of Genetics, 38, 645679.
Pennisi E (2009) Wanted: clean Neandertal DNA. Science, 323,
868.
Poinar HN, Schwarz C, Qi J et al. (2006) Metagenomics to pale-
ogenomics: large-scale sequencing of mammoth DNA. Science,
311, 392394.
Rohland N, Hofreiter M (2007a) Ancient DNA extraction from
bones and teeth. Nature Protocols, 2, 17561762.
Rohland N, Hofreiter M (2007b) Comparison and optimization
of ancient DNA extraction. BioTechniques, 42, 343352.
Schuster SC (2008) Next-generation sequencing transforms
todays biology. Nature Methods, 5, 1618.
 2009 Blackwell Publishing Ltd
T E C H N I C A L A D V A N C E S 683
Schwarz C, Debruyne R, Kuch M et al. (2009) New insights from
old bones: DNA preservation and degradation in permafrost
preserved mammoth remains. Nucleic Acids Research, 37, 3215
3229.
Shapiro B, Drummond AJ, Rambaut A et al. (2004) Rise and fall
of the Beringian steppe bison. Science, 306, 15611565.
Stiller M, Knapp M, Stenzel U, Hofreiter M, Meyer M (2009)
Direct multiplex sequencing (DMPS): a novel method for tar-
geted high-throughput sequencing of ancient and highly
degraded DNA. Genome Research, 19, 18431848.
Vigilant L, Hofreiter M, Siedel H, Boesch C (2001) Paternity and
relatedness in wild chimpanzee communities. Proceedings of
the National Academy of Sciences, USA, 98, 1289012895.
Wandeler P, Hoeck PE, Keller LF (2007) Back to the future:
museum specimens in population genetics. Trends in Ecology
& Evolution, 22, 634642.
Weinstock J, Willerslev E, Sher A et al. (2005) Evolution, system-
atics, and phylogeography of pleistocene horses in the new
world: a molecular perspective. PLoS Biology, 3, e241.
Supporting Information
Additional supporting information may be found in the online
version of this article.
Table S1 Cave bear samples used in this study.
Table S2a Results of qPCR comparing different amounts of
sample powder in the extraction buffer (kept constant at 5 mL).
Table S2b Results of a second qPCR using two of the extracts as
in Table S2a, but using further dilution steps comparing differ-
ent sample powder input and several dilutions.
Table S3a Results of qPCR comparing different ratios of extrac-
tion to binding buffer (amount of extraction buffer kept constant
at 5 mL).
Table S3b Results of qPCR comparing different ratios of extrac-
tion buffer to binding buffer (amount of extraction buffer kept
constant at 5 mL) reducing the amount of binding buffer even
further compared to Table S3a.
Table S4 Results of qPCR comparing the initial extraction tech-
nique with the new column-based extraction method on nine
cave bear samples.
Table S5 Results for the chimpanzee samples.
Please note: Wiley-Blackwell are not responsible for the content
or functionality of any supporting information supplied by the
authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.